Anysearch MCP Server

AnySearch MCP Server Skill

Skill by ara.so — MCP Skills collection.

AnySearch MCP Server is a unified real-time search engine for AI agents, providing general web search, 23 vertical domain searches (finance, academic, security, legal, code, etc.), parallel batch search, and full-page URL content extraction via the MCP protocol.

What It Does

• General Web Search: Natural language queries across the web
• Vertical Domain Search: Structured queries in 23 specialized domains
• Batch Search: Execute up to 5 independent queries in parallel
• URL Extraction: Fetch and convert full page content to Markdown
• Anonymous Access: Works without API key (lower rate limits)

Installation

Streamable HTTP (Recommended - No Proxy)

For agents supporting MCP spec 2025-03-26+ (OpenCode, Claude Desktop 2025.6+):

OpenCode (~/.opencode/config.json or opencode.json):

{
  "mcp": {
    "anysearch": {
      "type": "streamable-http",
      "url": "https://api.anysearch.com/mcp",
      "headers": {
        "Authorization": "Bearer ${ANYSEARCH_API_KEY}"
      }
    }
  }
}

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "anysearch": {
      "type": "streamable-http",
      "url": "https://api.anysearch.com/mcp",
      "headers": {
        "Authorization": "Bearer ${ANYSEARCH_API_KEY}"
      }
    }
  }
}

stdio Transport (Via mcp-remote Proxy)

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "anysearch": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.anysearch.com/mcp",
        "--header",
        "Authorization: Bearer ${ANYSEARCH_API_KEY}"
      ]
    }
  }
}

VS Code Copilot (.vscode/mcp.json):

{
  "servers": {
    "anysearch": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.anysearch.com/mcp",
        "--header",
        "Authorization: Bearer ${ANYSEARCH_API_KEY}"
      ]
    }
  }
}

Cline (VS Code settings):

{
  "mcpServers": {
    "anysearch": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.anysearch.com/mcp",
        "--header",
        "Authorization: Bearer ${ANYSEARCH_API_KEY}"
      ]
    }
  }
}

SSE Transport (Via supergateway Proxy)

Start the proxy server:

npx -y supergateway \
  --streamableHttp https://api.anysearch.com/mcp \
  --outputTransport sse \
  --port 8000 \
  --oauth2Bearer ${ANYSEARCH_API_KEY}

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "anysearch": {
      "type": "sse",
      "url": "http://localhost:8000/sse"
    }
  }
}

Windsurf (~/.codeium/windsurf/mcp_config.json):

{
  "mcpServers": {
    "anysearch": {
      "serverUrl": "http://localhost:8000/sse"
    }
  }
}

API Key Configuration

Optional but recommended. Get a free key at: https://anysearch.com/console/api-keys

Priority order:

1. Authorization header / --api_key flag
2. ANYSEARCH_API_KEY environment variable
3. .env file with ANYSEARCH_API_KEY=<key>
4. Anonymous access (lower rate limits)

Set environment variable:

export ANYSEARCH_API_KEY="your-api-key-here"

Or create .env file:

ANYSEARCH_API_KEY=your-api-key-here

Available Tools

search - General or Vertical Domain Search

Execute a search query. For vertical searches, you MUST call list_domains first.

Parameters:

General Web Search Example:

{
  "query": "quantum computing breakthroughs 2025",
  "max_results": 5,
  "freshness": "month"
}

Vertical Search Example (Stock):

First, call list_domains:

{
  "domain": "finance"
}

Then use the discovered sub_domain and query_format:

{
  "query": "AAPL",
  "domain": "finance",
  "sub_domain": "finance.us_stock",
  "max_results": 5
}

News Search Example:

{
  "query": "artificial intelligence",
  "content_types": ["news"],
  "freshness": "week",
  "max_results": 10
}

Academic Search Example:

{
  "query": "10.1038/s41586-019-1666-5",
  "domain": "academic",
  "sub_domain": "academic.doi"
}

list_domains - Query Vertical Domain Directory

MUST be called before vertical search to discover available sub_domains and their query formats.

Parameters:

Single Domain Example:

{
  "domain": "finance"
}

Multiple Domains Example:

{
  "domains": ["finance", "academic", "security"]
}

Returns Markdown table with:

• sub_domain: Routing key for search
• description: What the sub_domain searches
• query_format: Exact format required (e.g., raw ticker, DOI)
• params_schema: JSON schema for optional parameters
• zone: If CN, must set zone: "cn" in search

batch_search - Parallel Search Execution

Execute 2-5 independent search queries in parallel. Single failure doesn't block others.

Parameters:

Example:

{
  "queries": [
    {
      "query": "AAPL",
      "domain": "finance",
      "sub_domain": "finance.us_stock"
    },
    {
      "query": "python async http client",
      "domain": "code",
      "sub_domain": "code.general"
    },
    {
      "query": "CVE-2024-1234",
      "domain": "security",
      "sub_domain": "security.cve"
    }
  ]
}

extract - URL Content Extraction

Fetch full page content from a URL and return as Markdown. HTML pages only, truncated at 50,000 characters.

Parameters:

Example:

{
  "url": "https://en.wikipedia.org/wiki/Quantum_computing"
}

Supported Domains

23 vertical domains available:

• code - Code repositories, packages, documentation
• tech - Technology news, products, reviews
• fashion - Fashion trends, brands, products
• travel - Flights, hotels, destinations
• home - Home improvement, furniture, decor
• ecommerce - Products, prices, reviews
• gaming - Games, platforms, news
• film - Movies, reviews, showtimes
• music - Songs, artists, albums
• finance - Stocks, crypto, markets, companies
• academic - Papers, patents, research
• legal - Laws, regulations, cases
• business - Companies, industries, markets
• ip - Patents, trademarks, copyrights
• security - CVEs, vulnerabilities, threats
• education - Courses, schools, materials
• health - Medical, wellness, drugs
• religion - Religious texts, organizations
• geo - Geographic data, maps, places
• environment - Climate, conservation, ecology
• energy - Energy sources, efficiency, policy
• ugc - User-generated content, forums

Common Usage Patterns

Pattern 1: General Web Search

When user asks an open-ended question:

{
  "query": "best practices for microservices architecture",
  "max_results": 10
}

Pattern 2: Recent News

When user wants current information:

{
  "query": "SpaceX launches",
  "content_types": ["news"],
  "freshness": "week",
  "max_results": 5
}

Pattern 3: Vertical Search Workflow

When user mentions structured identifiers (stock ticker, CVE, DOI, etc.):

Step 1: Identify the domain and call list_domains

{
  "domain": "security"
}

Step 2: Parse the response to find the correct sub_domain and query_format

Step 3: Execute search with exact format

{
  "query": "CVE-2024-1234",
  "domain": "security",
  "sub_domain": "security.cve"
}

Pattern 4: Multi-Intent Queries

When user has multiple independent questions:

{
  "queries": [
    {
      "query": "TSLA",
      "domain": "finance",
      "sub_domain": "finance.us_stock"
    },
    {
      "query": "electric vehicle market trends",
      "freshness": "month"
    }
  ]
}

Pattern 5: Deep Content Extraction

When search results provide URLs but user needs full content:

Step 1: Search

{
  "query": "kubernetes best practices",
  "max_results": 3
}

Step 2: Extract top result

{
  "url": "https://kubernetes.io/docs/concepts/configuration/overview/"
}

Pattern 6: Code Search

When user needs code examples:

{
  "query": "react hooks useEffect",
  "domain": "code",
  "sub_domain": "code.general",
  "max_results": 10
}

Pattern 7: Academic Research

When user references DOI or research topic:

{
  "query": "10.1038/nature12373",
  "domain": "academic",
  "sub_domain": "academic.doi"
}

Decision Flow

Use this logic to choose the right tool:

User query
  |
  +-- Has structured identifiers? (ticker, CVE, DOI, IATA, patent, etc.)
  |     YES -> 1) list_domains(domain) -> discover sub_domain & query_format
  |             2) search with domain + sub_domain + exact format
  |
  +-- Multiple independent questions?
  |     YES -> batch_search with array of queries
  |
  +-- Need full article/page content beyond snippets?
  |     YES -> search -> extract(url) from results
  |
  +-- Filter by content type or freshness?
  |     YES -> search with content_types and/or freshness params
  |
  +-- Otherwise -> search (general, natural language)

Vertical Search Requirements

Before executing vertical search, you MUST:

1. Call list_domains for the target domain
2. Follow query_format exactly (e.g., raw ticker "AAPL", not "Apple stock")
3. Use correct sub_domain that matches user intent
4. Set zone: "cn" if the sub_domain requires it (check zone field)
5. Include sub_domain_params if the schema requires additional fields

Troubleshooting

Issue: "Invalid query format for vertical search"

Cause: Query doesn't match the query_format from list_domains

Solution:

1. Call list_domains for the domain
2. Check the query_format field
3. Transform user input to match exact format (e.g., extract ticker symbol from natural language)

Issue: "Missing required zone parameter"

Cause: Sub_domain requires zone: "cn" but it wasn't provided

Solution:

1. Check list_domains response for zone field
2. If zone: "CN", add "zone": "cn" to search call

Issue: "Anonymous rate limit exceeded"

Cause: No API key configured, hitting lower anonymous limits

Solution:

1. Get free API key from https://anysearch.com/console/api-keys
2. Set ANYSEARCH_API_KEY environment variable
3. Restart agent to pick up new key

Issue: "SSE/stdio proxy not working"

Cause: Proxy server not running or incorrect configuration

Solution for SSE:

1. Ensure supergateway proxy is running: npx -y supergateway --streamableHttp https://api.anysearch.com/mcp --outputTransport sse --port 8000 --oauth2Bearer ${ANYSEARCH_API_KEY}
2. Check proxy is accessible: curl http://localhost:8000/sse
3. Verify agent config points to http://localhost:8000/sse

Solution for stdio:

1. Use mcp-remote proxy (simpler, auto-detects transport)
2. Verify npx can download packages
3. Check agent restart picked up new config

Issue: "No results for vertical search"

Cause: Wrong sub_domain selected or malformed query

Solution:

1. Review list_domains output again
2. Match user intent to description field of sub_domains
3. Verify query matches query_format exactly
4. Try general search first to validate the entity exists

Issue: "Extract returns truncated content"

Cause: Page content exceeds 50,000 character limit

Solution:

1. This is expected behavior for very long pages
2. Use search results snippets for overview
3. Extract multiple specific URLs if needed
4. For extremely long docs, consider asking user which sections to prioritize

Configuration Quick Reference

Best Practices

1. Always call list_domains before vertical search - Don't assume you know the sub_domains
2. Use batch_search for multiple intents - More efficient than sequential calls
3. Prefer general search for exploratory queries - Vertical search requires exact formats
4. Extract URLs sparingly - Search snippets are often sufficient
5. Set appropriate max_results - Default 10 is good, reduce for speed, increase for thoroughness
6. Use freshness for time-sensitive queries - News, events, recent developments
7. Never include API keys in chat - Always use environment variables
8. Check zone requirements - CN sub_domains require zone: "cn" parameter

Security Notes

• All queries and URLs are sent to https://api.anysearch.com
• Do not use for sensitive data (passwords, personal info, trade secrets)
• API keys grant higher rate limits - protect them as secrets
• Anonymous access works but has lower limits